We analyzed more than 15,000 seasoned Seed investments made on AngelList, and one clear lesson stood out: writing smaller checks (presumably into investment rounds that forced GPs into smaller allocations) outperformed rounds where GPs could take their typical allocations. The data also suggest there’s no consistent “skill” in check sizing. Taken together, these findings paint a picture of early-stage venture capital as an efficient, highly competitive market where access matters more than perfect precision.
Why should a GP want to write different-sized checks?
Quantitative finance began as a discipline by studying the portfolio allocation problem, or more simply, what size checks to write into investment opportunities of different perceived quality. Taking two simplifying assumptions that probably hold for Seed investing:
That the idiosyncratic returns of Seed investments are uncorrelated, and
That every Seed investment has equal volatility
It follows that GPs would achieve their optimal risk-reward portfolio by investing in proportion to the “signal” they observe for each startup. These are simplifying assumptions made for math. But the key insight is that, because a GP's investable capital is limited, they might do well to invest more in their most promising opportunities and less in their less promising opportunities.
This is a stylized model that doesn’t take into account the realities of running a venture fund. In practice, writing a small check takes almost as much time and effort as writing a large one — but a large check, if successful, will return much more money to the fund. From a fund-level perspective, that means a GP should prefer to invest as much as possible in any company they believe is worth backing.
Of course, most GPs can’t do this freely — they’re constrained by the commitments they made to LPs about check size, diversification, or maximum position limits. But the key insight is this: in a fund-level framework, we get a different dynamic than what would be predicted by quantitative finance. GPs almost never choose to write small checks — they write them when allocation is scarce. Small checks may therefore represent high-conviction situations where the GP wanted to commit more capital but couldn’t.
For this analysis, we only consider Seed investments made through 2022 on the AngelList platform that were priced and raised in USD. We only consider the first investment that a GP made into a startup, not follow-on bridge rounds.
Given the evolution of Seed-stage venture capital and the AngelList platform over the past 15 years, we generally work with annual buckets of investments to correct for any time effects. As a result, GPs must have made five Seed investments in a given year for those investments to be included in our analysis.
To accurately include investments that GPs made through funds, we use gross numbers (i.e., multiple on invested capital, “MOIC”), which do not include the fees and carry that investors may pay. Investment performance was captured as of July 1, 2025.
Do GPs size checks?
Yes, GPs typically write differently sized checks based on their conviction or allocation. This is as opposed to writing similar-sized checks for each opportunity or sizing based on valuation to get a consistent ownership percentage.
To gain a fuller understanding, we group GPs by year and filter down into the years they made at least five Seed investments. We then compare the median check written by each GP each year to the largest and smallest checks we observed each year. We obtained the following frequencies from 933 observations:
Two-thirds of observations saw GPs go outside the (fairly wide) 0.5–2x median band for their annual investments. In other words, if a GP’s median Seed investment in a given year was $100k, two-thirds of the time they would write a check smaller than $50k or larger than $200k into a Seed-stage startup that same year.
We also asked whether check sizing is related to pre-money valuation (which often corresponds to larger fundraising rounds). Here, the data show only a lightly positive correlation of 0.10. This suggests that GP check sizing isn’t being driven by a rigid ownership percentage target. Our view is that ownership percentage may become a bigger consideration for investors in later rounds; however, for this analysis, we focused exclusively on Seed investments.
Is there a correlation between check size and higher returns?
Large checks tend to perform slightly better than “typical” ones on average, but the signal is weak. In fact, it’s small checks that perform best—likely because GPs get squeezed down in the most competitive, highest-quality deals. This aligns with our prior perspective on how small checks necessarily must provide higher returns to make them valuable from a fund-level perspective.
To dive deeper here, we divided the more than 15,000 Seed checks from our GP-year filter above into three buckets:
Typical: Checks between 0.5x and 2x of the GP’s median Seed check that year
Small: Checks smaller than half the GP’s median Seed check that year
Big: Checks more than twice the size of the GP’s median Seed check that year
As would be expected, most of the Seed checks we looked at (about 80%) fell in the first, “typical” bucket. The following table shows the count, average multiple, 75th percentile multiple, and fraction of investments with a multiple of less than 0.9x (”money losers”).
It’s important to note that general power-law rules apply here: investment count can be a major factor in average performance (“make more investments, get higher average returns”). That’s why we include the 75th percentile MOIC (multiple on invested capital)—to add an additional check on how much of average outperformance is being driven by one or two outliers.
From the data, two key takeaways stand out:
Small checks were the best performers. Despite being the smallest category, they delivered the highest average returns (outperforming typical checks by 1.17x) and the strongest 75th percentile returns—even though they had slightly more “losers”.
Large checks showed slight outperformance. On average, large checks beat typical-sized checks by just 1%, making the signal less clear.
From conversations with GPs about their small checks, we learned that the most common reason a GP would write a small check is because the interest of other investors results in them being “crammed down,” not a lack of conviction. It appears that the common signal shared across investors about a startup is more potentially important than the idiosyncratic signal that an individual investor has about that startup.
Smaller checks may show higher multiples on invested capital on average, but that doesn’t mean they’re “better.” VCs (and their LP investors) ultimately care about performance on a fund level, and a small check that triples won’t move the needle on returning a fund. What’s more interesting is the structural dynamic—because it’s harder for smaller checks to matter at the fund level, VCs are reluctant to write them. That reluctance creates space where outperformance can persist, even in an efficient venture market.
On the other hand, GPs only write large checks for high-conviction opportunities. However, there may be a “winner’s curse” effect here for large checks. A GP’s private signal about a startup that makes them want to write a larger check could be offset by the very fact that they were able to write a larger check into the round, which suggests that other investors may not be so keen on the startup.
Finally, check size correlates with downside risk: big checks had the fewest losers, while small checks had the most. Since the most common way a Seed investment fails is by running out of runway and shutting down, this could reflect a structural effect—smaller checks often go into smaller rounds with less runway, while larger checks tend to be part of bigger rounds with more runway.
Is there consistency among GPs in check-size skill?
We can estimate how good a GP is at check sizing by randomly permuting the checks they actually wrote into companies and comparing the return of that permuted portfolio to the return of their actual portfolio. We call the percentile at which their real portfolio stands against 1,000 simulated shuffled portfolios the GP’s “Check-Sizing Portfolio Percentile.” A GP with a percentile greater than 50 shows “alpha” in check sizing, since their real portfolio performed better than the typical shuffle.
If check sizing exists, it should be replicable across time periods. To test for persistence of this signal, we split time into two buckets: Seed investments made prior to 2020 and those made after (recall that 2022-vintage investments are the youngest investments we’re considering). We also restricted the analysis to GPs who made at least ten Seed investmentsin each period, leaving us with a pool of 56 GPs.
The results?
No identifiable alpha in check sizing. Average and median percentiles came in below 50 in both time periods, showing that the “small checks do better” effect dominates our results.
No persistence over time. Performance didn’t carry forward across periods—the correlation was -0.12, and only 3 out of 56 GPs stayed above the 60th percentile in both.
The conclusion is clear: there’s no consistent ability for individual GPs to size checks skillfully. Instead, the pattern that stands out is the “competitive deals are the best, and you can only get a smaller check in” phenomenon.
Do Small Checks Even Matter?
Smaller checks may show higher multiples on invested capital on average, but that doesn’t mean they’re “better.” VCs (and their LP investors) ultimately care about performance on a fund level, and a small check that triples won’t move the needle on returning a fund. What’s more interesting is the structural dynamic—because it’s harder for smaller checks to matter at the fund level, VCs are reluctant to write them. That reluctance creates space where outperformance can persist, even in an efficient venture market.
What does this tell us about Seed investing as an asset class?
Broadly speaking, there are three models of how venture outcomes might work:
The “investment” model: some GPs skillfully identify undervalued, non-consensus opportunities (similar to how hedge funds aim to operate in public markets).
The “access” model: the real competition is for GPs to gain access to the best startups, as the ability to assess a deal is broadly shared knowledge.
The “chaos” model: nobody knows anything, and outcomes are largely random.
Our findings point most strongly to the “access” model, with a healthy dose of “chaos.” Small checks consistently outperform, suggesting they’re written in competitive rounds where access is scarce—exactly what the “access” model predicts. At the same time, we found no replicable ability in investment sizing across time periods, which runs counter to the “investment” model. The only modest support for the “investment” model comes from the slight outperformance of large checks over typical ones, which appears to be a population-level effect not observable or replicable on the level of individual GPs.
The potential practical implications of these findings are:
For GPs: GPs may be incentivized to fight hard for allocation in competitive deals, regardless of final check size. The data shows that those small allocations are often the highest-performing investments - which is why they could still make sense even if they generate lower returns on an absolute basis.
For LPs evaluating track records: Our data suggests that check sizing skills don’t reliably replicate at the individual GP level.
For LPs managing exposure:
If you’re exposed to many GPs, large checks aren’t a bad idea at the portfolio level— even if it’s impossible to predict which GP will hit big with one.
If you’re exposed to very few GPs, increased check-sizing discipline from those GPs may help reduce variance and avoid outsized risks tied to conviction bets.
[Bonus] Style Drift and Adverse Selection
It’s widely believed that style drift is one of the menaces of GP selection. Essentially, LPs are cautious of GPs trying new ideas, particularly when the LPs decided to back a GP based on their historical track record.
One of the interesting implications of our analysis is that the risk of style drift may not be from a GP trying new ideas, but rather from adverse selection on deals of that type. Imagine an investor known for consumer deals moves into doing climate-tech deals. The risk to that change may simply be that the climate-tech deals the investor sees will be ones that have already been passed on by more famous climate-tech investors.
Disclaimer
This presentation and the information contained herein is provided for informational and discussion purposes only and is not intended to be a recommendation for any investment or other financial, legal, or tax advice of any kind, and shall not constitute or imply any offer to purchase, sell, or hold any security or to enter into or engage in any type of transaction. Any such offers will only be made pursuant to formal offering materials containing full details regarding risks, minimum investment, fees, and expenses of such transaction. This document and the information, charts, and graphs provided within are for informational purposes solely and should not be relied upon when making any investment decision. There is no guarantee that the fund will achieve the same exposure to or quality of the companies discussed herein. Data used in this presentation comes from AngelList and was current as of July 1, 2025.
